With the rapidly increasing reliance of life sciences research on digital data and sophisticated computational analysis techniques, the ability to reproduce findings generated by in-silico data analysis workflows is of increasing importance to the scientific community. However, challenges to reproducibility arise from the heterogeneity of approaches available to workflow definition and enactment. In this work, we explore the requirements of ‘digital reproducibility’ of genomic data analysis by implementing a complex but extensively used bioinformatics workflow using three widely adopted approaches that differ markedly in their implementation. We use these case studies to identify implicit and explicit assumptions present in each of the workflow implementations and then propose a comprehensive breakdown of reproducibility requirements that can inform a more structured, assumption-free approach to implementing highly reproducible genomic analyses.